ACS Data Users Group

Hello All,

Does anyone have documentation discussing the homeless and poverty levels? Are the homeless included in the "poverty universe?" Is it possible to identify homeless counts? I am searching PUMS data and can also IPUMS data.

Specifically, I'm attempting to answer the question, "If the number of people in Washington state under 200% FPL is decreasing, why is the homeless population increasing?" Yes, I understand there is not a direct correlation between homelessness and poverty... but this is where I am starting.

Any ideas?

Thank you,

Lorna

Homeless data in general is fraught, and in ACS especially. There's a good discussion here: acsdatacommunity.prb.org/.../homeless-population-and-those-in-transitional-housing

Yes, I went through that discussion. Thank you though. I was just hoping there has been an update since then.

Most group quarters population (such as those living in dorms, barracks, correctional facilities, nursing homes) are not in the universe for whom poverty rates are calculated. If someone has low income and is housed, they're in the universe. If someone has low income and is unhoused, they may not in the universe (or survey at all--an often missed population). Only noninstitutional group quarters "not elsewhere classified" are included. (There are also challenges tracking the population who are "doubled up." They would be included in the universe for whom poverty rates are calculated, but are sometimes missed in surveys.)

This makes a lot of sense. Thank you!!

To Lorna and Bea--

They do get income data from those living in group quarters, but shelters, etc. are collected differently. I

n terms of fining the homeless, we worked with the NYtimes for a project on homeless in LA and the astounding thing we found was that if one used PUMAs and PUMS, we found that the areas where the proportion of income spent on housing (rent) went toward the moon, I think it can go as high as 95% were the same areas (comparing data from a decade or so prior with more current) as where the homelessness were known to have increased. So sometimes you can find things out when you are willing to look at things like that.

Andy

Dear Lorna,

At one point I was working on stats for the homeless population. Homeless shelters, food trucks etc are part of the decennial census. I don't think that the ACS collects group quarters data on these people. Beth is an expert on this (she has written papers and conference proceedings)

At some point I volunteered with an organization (cooking and serving meals on Sunday) that has formally homeless people who are residents, Many have a history of alcohol and drug abuse. I don't know if the ACS surveys this population but I think that they do but the group quarters for the ACS is institutional/non-institutional The residents are served meals in a cafeteria. There are no kitchens in the room so the resident population would be a group quarters. The organization lets in homeless people off of the streets every night. The receive a hot meal and and sleep on mats on floor that are put out every night. During the day they are out on the street. A list of group quarters codes are here https://portal.ct.gov/-/media/Departments-and-Agencies/DPH/Population/Group-Quarters/2020-GQ-Definitions-and-Code-List.pdf

In Boston the Pine Street Inn that is the largest organization in the Boston Area. that works with homeless populations has a "systems" group. I was in contact with them at some point. That group has statistical models for the dynamics of the local homeless population. I can ask what type of data they use.

There is a census web page with a list of the 2020 decennial census codes but I have a hard time finding the web page.

Here is a blog post:

https://www.census.gov/newsroom/blogs/random-samplings/2021/03/2020-census-group-quarters.html

You might look at P5 and P18 but they don't break out homeless group quarters codes.

If you go back to the 2010 decennial census there is a finer breakout.

PCT39 GROUP QUARTERS POPULATION BY SEX BY AGE BY GROUP QUARTERS TYPE

For 2020 decennial census there is for example

code 702 Soup KItchens

Code: 706 Targeted Non-Sheltered Outdoor Locations

There is the HUD Point in time annual census:

https://www.hud.gov/press/press_releases_media_advisories/hud_no_23_278

Only certain city/towns collect this data. For example in the Boston Area Boston MA and Cambridge MA

You might be able to get a custom tabulation for the 2020 Decennial census with codes 702 and 706 broken for along with covariates for some type of geography. I looked into this. The minimum charge for a custom tabulation is about $3000. but if you want a custom tabulation Dorer Community Service Foundation would be willing to split the cost. Once there is a custom tabulation it is available to the public.

Additional geographies are free so you could get the entire US. You might be able to get tracts broken out.

PCT39 for 2010 has

Emergency and transitional shelters (with sleeping facilities) for people experiencing homelessness (701)

broken out.

If you can get a custom tabulations with some covariates (PCT39 has age and sex as covariates) you could use the technique used in my Small Area estimation program to adjust (using ACS 5 year tables) to years after 2020.

there is the American Housing Survey (AHS) -- I think that the Bureau of Census does the survey for HUD.

The PUMS datasets are here:

https://www.census.gov/programs-surveys/ahs/data/2021/ahs-2021-public-use-file--puf-/2021-ahs-metropolitan-puf-microdata.html

I think that the geography is the "Metropolitan Area". Again you can try to use ACS detail tables, PUMS file and the Small Area Estimation package to get data for census tracts and 2022. You will need to do a lot of "hand waiving" to try to come up with some type of validation.

Since I have your email, we can continue this discussion via email or zoom.

Best

Dave Dorer

Here is a paper on the homeless population and a statistical model:

www.ncbi.nlm.nih.gov/.../pone.0284336.pdf

When I have a minute I'll look through the paper to see what data sources they used.

Just looked They used data from

Homeless Management Information System (HUD)

https://www.hudexchange.info/programs/hmis/

I don't know how they got the data. You may need to go through an Institutional Review Board (IRB). Abt was on the paper and they have an IRB. I've worked with ABT on a project that used confidential SEER Medicare data.

Since you work for the Washington DSHS you should be able to get access to the data. I don't know if they send out de-identified datasets but I am qualified to de-identify a dataset under HIPPA regulations (see DCSF website)

https://dorerfoundation.org/documents/

Dear Lorna,

For the 2020 census there is something called Privacy Protected Microdata Files (PPMF) I have a message in to find out the release schedule and technical documentation. You might be able to get group quarters type age race and sex but I'm not sure. You could then link the data to the B01001A-G and B17001 tables for blockgroups and tracts to get the relationship between the homeless counts from the PPMF and poverty for the associated blockgroup. You can also adjust to 2022 using ACS 2022 tables. Stay tuned.

Best,

Dave

Dear Lorna,

Here is a model from HUD:

https://www.huduser.gov/portal/sites/default/files/pdf/Market-Predictors-of-Homelessness.pdf

https://www.huduser.gov/portal/datasets/hpmd.html

https://www.hudexchange.info/resource/3031/pit-and-hic-data-since-2007/

It combines data from the HUD PIT and CoC datasets along with ACS data. for example Employment, Poverty (US Census version) and other sources. This might point you in the right direction as far as how to estimate the relationship between Poverty and Homelessness based on ACS data.

Dave